This article is the third in a three-part series about cycling in the heatwaves that are becoming ever more frequent as we experience climate change. Part 1 explored the physics of thermoregulation. Part 2 was about how the body copes when riding in high temperatures. This article covers physiological changes that occur after several days of heat exposure.
https://science4performance.github.io/RidingTheHeatwave3
You might also enjoy my article about fuelling.
This article is the second in a three-part series about cycling in the heatwaves that are becoming ever more frequent as we experience climate change. Part 1 explored the physics of thermoregulation. This article is about how the body copes when riding in high temperatures. Part 3 covers physiological changes that occur after several days of heat exposure.
This article is the first in a three-part series about cycling in the heatwaves that are becoming ever more frequent as we experience climate change. We explore the physics of thermoregulation. Part 2 is about how the body copes when riding in high temperatures. Part 3 covers physiological changes that occur after several days of heat exposure.
Have we reached the stage where agentic AI gives you superpowers? We have moved on from chatbots to AI agents that can carry out tasks for you. The most powerful way to use an agentic AI is to Augment your Intelligence. I set myself a goal of completing a substantial project on a Saturday afternoon.
Even the free versions of AIs will do amazing things, but if you want to create something useful, you need to have a vision of what you’d like to achieve. A good starting place is to consider your own personal areas of expertise. I decided to use the 81 blogs on this web site as my subject matter. My vision was to generate the 3D semantic map that appears at the top of this page.
AI tools: I decided to use the Copilot agent in VSCode to perform web scraping, coding and data processing tasks. Gemini provided some useful pointers along the way. I also wanted to experiment with running a language model locally.
Data: Scrape all the blogs from my web site
Processing: Create a semantic representation of each blog in the form of an embedding. Reduce the dimensionality of the embeddings. Look for clusters. Identify common semantic characteristics of the members of each cluster.
Presentation: Turn it into a 3D plot, highlighting the clusters and the progression of time.
I asked Copilot to create a uv python environment in VSCode. In order to scrape all my blogs, Copilot first found the sitemap.xlm and then I asked it to create a JSON file with the page name, date and text of each blog. After a little tidying up this was all done in about half an hour.
I turned to Gemini for suggestions of the best publicly-available text embedding model on HuggingFace that would run on my MacBook Pro. It proposed a Qwen2-7B model due to its large context window and strong performance. I had already downloaded LM Studio, so it is was simple (or so I thought) to download the model and set it running on a local server. However, after numerous attempts, Copilot could not get a response from the server, even though LM Studio confirmed it was running. Eventually Gemini suggested that the Qwen2 model was running as a chatbot rather than an embedding domain. I eventually found the “Override Domain Type” in the LM Studio models tab. Once I switched it to Embedding, Bingo! Everything worked.
Copilot’s Python code successfully created text embeddings for all the blogs. My intuition was that blogs on similar topics would be closer to each other in embedding space, but the embedding dimension was 3584. Fortunately, along with recommending a text embedding model, Gemini had recommended using UMAP, as the gold standard in visualising complex data sets while preserving clusters, in preference to alternatives, such as Principal Components Analysis, and to use Plotly for visualisation.
Copilot wrote scripts to collapse the dimensions to three and to display the results in a 3D Plotly chart. This all ran perfectly. Upon inspection, I could see distinct groupings, so I asked Copilot to identify four clusters. It decided to use K-means, which was fine by me, but this unsupervised method doesn’t explain why the points in the same cluster were close to each other. So I went back to Copilot and instructed it to review to the original 3584 dimensional embeddings and identify, for each of the four clusters, what aspects of the text explain why the blogs have been grouped together?
Copilot chugged away for a while and came back with a very nice characterisation of each group. It is instructive to look at the code to find out how it did this. The code shifted back into the full embedding space and identified the five blogs closest to the centroid of each cluster. It appeared to use snippets of the first 220 characters of the top five blogs to identify the themes. I think this worked because each blog opens with a summary of the topic. Copilot produced the short labels for the clusters that appear on the chart.
I also asked Copilot to attempt to find a semantic interpretation of the three axes that resulted from the UMAP projection.
An interactive 3D chart provided an intuitive visualisation of the results. I asked Copilot to set the marker shape according to cluster and to colour the points according to date to see whether themes have changed over time.
The blogs on this web site are characterised by four themes
The themes vary in nature in the following manner
GitHub repository for this project https://github.com/science4performance/Science4PerformanceWebsite
A high-performance bicycle relies of a mix of advanced polymer science and metallurgy. Carbon only makes up about the half the mass of a carbon-framed bike. The moving components include iron, aluminium and specialist alloys. The frame is held together with epoxy resins and the tyres include a range of compounds to reduce rolling resistance, while maintaining grip. I wondered, what is the chemical composition of my bicycle? Where do these chemicals come from?
The primary materials of the frame and rims are carbon fibre and epoxy resin. High-end frames and rims use “pre-preg” carbon filaments held together by a thermosetting resin matrix. Carbon fibre is roughly 95% elemental carbon. It is created by heating precursor fibres until only carbon remains in a hexagonal crystalline structure. Epoxy resin is polymer typically derived from carbon, hydrogen and oxygen. It provides the compressive strength that keeps the carbon fibres in shape.
Estimated Mass: ~2.8 kg (Frame, Fork, Rims).
Elemental makeup: ~80% Carbon, ~12% Oxygen, ~7% Hydrogen, ~1% Nitrogen.
The drivetrain is made out of aluminium alloys, stainless steel and small amounts of titanium/chrome. The cranks and hubs require high-strength aluminium alloys that include zinc and magnesium to prevent fatigue. The cassette and chain are mostly chromium-steel. The chain requires high tensile strength and wear resistance, achieved through iron alloyed with carbon and chromium. Bearings are steel (iron/chromium) or occasionally ceramic (silicon nitride).
Estimated Mass: ~2.6 kg.
Elemental makeup: ~65% Iron, ~30% Aluminium, ~3% Chromium, ~2% Zinc/Magnesium/Others.
Tyres are made from synthetic/natural rubber, silica and carbon black, a reinforcing agent. The GP5000 is famous for its “Black Chili” compound. Unlike older tyres that relied heavily on carbon black, modern high-performance tyres use a high percentage of silica to reduce rolling resistance while maintaining grip. The casing is usually nylon (polyamide), consisting of carbon, nitrogen, oxygen and hydrogen. The bead is often Kevlar (aramid), which is another nitrogen-rich polymer.
Estimated Mass: ~0.5 kg (for the pair).
Elemental makeup: ~60% Carbon, ~20% Silicon, ~10% Oxygen, ~5% Sulphur (used in vulcanization), ~5% Hydrogen/Nitrogen
This table estimates the elemental distribution for a complete 8,000g (8kg) bike. These figures are calculated based on the average weight of the components listed above.
| Element | Estimated Mass (g) | % of Total | Primary Source |
| Carbon | 3,840g | 48.0% | Frame, wheels, tyres, resins, saddle |
| Iron | 1,760g | 22.0% | Chain, cassette, spokes, bearings, bolts |
| Aluminium | 1,440g | 18.0% | Crankset, hubs, stem, bars, calipers |
| Oxygen | 400g | 5.0% | Epoxy resins, rubber compounds, paint |
| Hydrogen | 240g | 3.0% | Polymer chains in resins and plastics |
| Silicon | 120g | 1.5% | Tyre compound (Silica), lubricants |
| Chromium | 80g | 1.0% | Stainless steel hardening (Drivetrain) |
| Nitrogen | 40g | 0.5% | Nylon tyre casing, Kevlar beads |
| Sulphur | 40g | 0.5% | Vulcanising agent in tyres and tubes |
| Others (Zn, Mg, Ti, Cu) | 40g | 0.5% | Aluminium alloying and specialty bolts |
| Total | 8,000g | 100% |
A fascinating paper by Craig Tindale, “The Return of Matter”, provides a sobering perspective on the dependency of manufacturers on the dirty and energy-intensive business of refining, purifying and separating the elements required for modern engineering and technology. While a Canyon bikes are designed in Germany and its Shimano components are engineered in Japan, the material reality of the bike is heavily dependent on Chinese industrial processing to turn the raw ore into high-purity metals and polymers.
Here is how this bike’s elemental components are tied to Chinese supply chains:
Component: Frame, Wheels, Resins.
Dependency: High.
While the article focuses on metals, it notes that China has spent decades building the “processing sovereignty” required for advanced materials. High-modulus carbon fibre and the epoxy resins that bind them are part of a complex polymer supply chain where China acts as a global gatekeeper. Even if the precursor chemicals are sourced elsewhere, the massive scale of carbon fibre “midstream” production is increasingly concentrated in China.
Component: Chain, Cassette, Spokes, Bearings.
Dependency: Total.
Tindale describes an “Iron Ore Stranglehold”. Even though Western majors like BHP and Rio Tinto mine the ore, it is shipped as concentrate directly to Chinese smelters. The steel in your Ultegra cassette is likely refined in a Chinese furnace that sets the global “tempo of Western inflation” and availability.
Component: Crankset, Hubs, Cockpit.
Dependency: Extreme (60% share).
China controls approximately 60% of global aluminium smelting. Furthermore, high-performance aluminium (like the 7000-series in your cranks) requires magnesium for hardening. China controls 90–95% of global magnesium smelting. Without Chinese magnesium, your bike’s aluminium components would lack the fatigue resistance necessary for racing.
Component: Tires (Silica), Lubricants.
Dependency: Dominant (95% share).
The refining of silicon is a massive Chinese monopoly; they control 95% of the world’s polysilicon capacity. While your tyres use silica , the high-purity chemical processing required for the “Black Chili” compound sits firmly behind what Tindale calls China’s “lattice of chemical plants”.
Component: Stainless steel, Alloying, Bolts.
Dependency: Structural (The “Derivative Mineral Trap”).
Titanium is used in high-end bolts and derailleur parts. China and Russia control 75% of global titanium sponge capacity. The US has only one domestic plant, leaving bike manufacturers with almost no non-adversarial choice for titanium. Chromium and other alloy ingredients are often recovered as “hitchhikers” during the smelting of host metals. Since China dominates base metal smelting (e.g., 50% of copper), it essentially “inherits” the critical by-products needed to harden your bike’s drivetrain.
You might “own” the bike, while Canyon and Shimano might “own” the design, but the kinetic power—the ability to actually build the machine—belongs to whoever owns the refineries. If China were to tighten export controls, as it has recently done for antimony (ammunition) and tungsten (munitions), the production of high-performance bikes would likely experience a “forced regression in engineering capabilities”, where manufacturers would have to substitute inferior, heavier materials for the refined ones they can no longer access.
A recent report in Science announced the publication of a new human blood protein atlas, describing the disease signatures of thousands of proteins circulating in the blood. Minimally invasive protein profiling marks a step forward in the personalisation of medicine. Some interesting statistical and machine learning techniques were employed.
The researchers’ methods included a technique called proximity extension assay (PEA), which makes use of highly specific probes of DNA strands to detect minute concentrations of proteins in the blood plasma. Amplification with PCR (Polymerase Chain Reactions) allowed 5,416 proteins to be evaluated.
A longitudinal dataset showed dramatic changes as children passed through adolescence to adulthood. The central part of the study was a cross-sectional analysis, where age, sex and BMI were identified as important explanatory factors. The signatures of 59 clinically relevant diseases, in seven classes, can be viewed interactively in The Human Protein Atlas.
Rather than the hideaway of a reclusive cockney, the secretome refers to the ensemble of secreted proteins. From a data science perspective, the challenge was how to find the signatures of a wide range of diseases, based on the differential abundance of over 5,400 proteins. This was complicated by the fact that many proteins elevated by a particular disease were also found to be elevated in other diseases.
“To investigate the distinct and shared proteomics signatures across diseases, we performed differential abundance analyses. Several groups were used as controls, including healthy samples, a disease background consisting of all other diseases, and samples from the same disease class.”
The differential abundance of proteins was evaluated using normalised protein expression units (NPX). The volcano chart above plots the p-values against the multiplicative (fold) change in NPX, both on log scales. The red values on the right were unusually high and the blue values on the left was exceptionally low.
The researchers used a logistic LASSO approach to identify the importance of proteins in providing a signature of each disease against its cohort. In the case of HIV above, CRTAM was the most significant explanatory factor, even though CD6 had the most extreme p-value.
A logistic model is trained on target values of one or zero, in this case representing the presence or absence of a disease. Least absolute shrinkage and selection operator (LASSO) is a version of linear regression that selects the most relevant explanatory variables using L1 regularisation. Adding the sum of the absolute values of the regression coefficients to the objective function forces the contribution of irrelevant variables towards zero as the hyper-parameter, λ, is increased. This property was particularly useful for the disease signature problem, where there were thousands of potential explanatory proteins.
The tricky aspect of LASSO is tuning the hyper-parameter, λ. You want it to be high enough to eliminate irrelevant variables, but not so high that it discounts the useful explanatory features. In the protein study, this was addressed using cross-validation: randomly splitting the data into 70:30 training and test sets, then rerunning the regression for a range of λ values. The quality of a model can be assessed in terms of both its accuracy and its required number of inputs, using criteria such as the Akaike information criterion or Bayesian information criterion, which favour parsimony. Repeating the randomisation 100 times, the researchers could home in on an optimal value of λ. The regression coefficients of the resulting model could then be used to rank the importance of the relevant proteins, as shown in the right hand side of the panel above.
The potential for a cheap, annual blood test to screen the whole population is immense. Proteomics adds to the arsenal of resources available to help people stay healthy. Early indications of diseases like cancer can be critical in initiating treatment. There is plenty of room to broaden the scope beyond the current 59 diseases, to include rarer conditions, such as Motor Neurone Disease, which has impacted some top sportsmen. It would be extremely helpful to find proteins related to the apparent epidemic of mental health issues, which are hard to define and lack objective, quantitative diagnostic criteria.
If you are a cyclist, athlete, dancer or exerciser struggling to reach your full potential, your might have a mismatch between your training and what you are eating. Persistently running an energy deficit can have an adverse impact on your health and performance, sometimes leading to a condition called Relative Energy Deficiency in Sport (REDs). Optimal training adaptations and peak achievements rely on consistently fuelling for the work required.
I have created an app that generates a score based on a short Personal Energy Availability Questionnaire (PEAQ) designed to identify people at risk.
The PEAQ is based on research published in BMJ Open Sport & Exercise Medicine, exploring the relationship between a REDs score derived from the questionnaire and quantified clinical consequences of low energy availability. A similar approach has been used in other research.
The app automates the scoring process and generates a free downloadable report that includes graphics and an interpretation of your result. It takes a few minutes to fill in your answers and the process is anonymous.
The report breaks down the overall score into three health categories. Physical health is based on body mass index (BMI) and injuries. Physiological factors include hormones, sleep and nutrition. Psychological wellbeing relates to habits and anxiety.
REDs is not confined to top athletes. It can occur in men and women of any age, at all levels of performance, across a spectrum of activities, including sports, exercise and dance.
Relative energy deficits can result from deliberate under-fuelling, particularly in activities where low body weight confers an aesthetic or performance advantage (dance, cycling, climbing, running etc.). Relative energy deficits can also arise, sometimes unintentionally, as a result of stepping up one’s training load without a corresponding increase in energy intake.
For evolutionary reasons, your body prioritises movement in the allocation of its energy budget. Energy availability is a measure of the amount of energy left over for day-to-day physiological processes: breathing, digestion, repair, brain function etc.. In an energy deficit, your body switches off inessential processes, such as reproduction. Poor bone health is one of the consequences of a reduction in sex steroid hormones. Other effects of low energy availability include fatigue, disrupted sleep and digestive problems.
For active people, low energy availability reduces your ability to perform high quality training/exercise and depletes your body’s ability to deliver the desired positive adaptations, such as muscle strength and endurance capacity.
Please take advantage of the PEAQ. If you have worries or concerns about your results, Dr Nicky Keay offers personalised health advisory appointments. You can find valuable resources at BASEM.
I built this educational health app in Python. It is hosted on the Streamlit Community Cloud. The code is on my GitHub page.
Mountjoy M, Ackerman KE, Bailey DM et al 2023 International Olympic Committee’s (IOC) consensus statement on Relative Energy Deficiency in Sport (REDs) British Journal of Sports Medicine 2023;57:1073-1098
Keay N Hormones, Health and Human Potential: A guide to understanding your hormones to optimise your health and performance, Sequoia books 2022
Keay N, Francis G, AusDancersOverseas Indicators and correlates of low energy availability in male and female dancers. BMJ Open in Sports and Exercise Medicine 2020
Nicolas J, Grafenuer S. Investigating pre-professional dancer health status and preventative health knowledge Front. Nutr. Sec. Sport and Exercise Nutrition. 2023 (10)
Keay N, Francis G. Longitudinal investigation of the range of adaptive responses of the female hormone network in pre- professional dancers in training March 2025 ResearchGate DOI: 10.13140/RG.2.2.30046.34880
Keay N. Current views on relative energy deficiency in sport (REDs). Focus Issue 6: Eating disorders. Cutting Edge Psychiatry in Practice CEPiP. 2024.1.98-102
Assessment of Relative Energy Deficiency in Sport, Malnutrition Prevalence in Female Endurance Runners by Energy Availability Questionnaire, Bioelectrical Impedance Analysis and Relationship with Ovulation status. Clinical Nutrition Open Science 2025S.
Sharp S, Keay N, Slee A. Body composition, malnutrition, and ovulation status as RED-S risk assessors in female endurance athletes, Clinical Nutrition ESPEN 2023, 58 :720-721
Keay N, Craghill E, Francis G Female Football Specific Energy Availability Questionnaire and Menstrual Cycle Hormone Monitoring. Sports Injr Med 2022; 6: 177
Nicola Keay, Martin Lanfear, Gavin Francis. Clinical application of monitoring indicators of female dancer health, including application of artificial intelligence in female hormone networks. Internal Journal of Sports Medicine and Rehabilitation, 2022; 5:24.
Nicola Keay, Martin Lanfear, Gavin Francis. Clinical application of interactive monitoring of indicators of health in professional dancers J Forensic Biomech, 2022, 12 (5) No:1000380
Keay, Francis, Hind Low energy availability assessed by a sport-specific questionnaire and clinical interview indicative of bone health, endocrine profile and cycling performance in competitive male cyclists BMJ Open Sports and Exercise Medicine 2018
Keay, Francis, Hind Clinical evaluation of education relating to nutrition and skeletal loading in competitive male road cyclists at risk of relative energy deficiency in sports (RED-S): 6-month randomised controlled trial BMJ Open Sports and Exercise Medicine 2019
Keay, Francis, Hind Bone health risk assessment in a clinical setting: an evaluation of a new screening tool for active populations MOJSports Medicine 2022;5(3):84-88. doi: 10.15406/mojsm.2022.05.00125″
The fascinating work of Geoffrey West explores the idea of universal scaling laws. He describes how the lifetimes of organisms tend to increase with size: elephants live longer than mice. On the other hand, average heart rate tends to decrease with size. It turns out that these two factors balance each other in such as way that over their lifetimes, elephants have roughly the same number of heartbeats as mice and all other animals: about 1.5 billion.
Less active people might be tempted to suggest that indulging in exercise reduces our lifetimes, because we use up our allocation of heartbeats more quickly. However, exercisers tend to have a lower resting heart rate than their sedentary peers. So if we really had a fixed allocation of heartbeats, would we be better off exercising or not?
To get a sense of how things change with scale, consider doubling the size of an object. Its surface area goes up 4 times (2 to the power of 2), while its volume and its mass rise 8 times (2 to the power of 3). Since an animal loses heat through its skin whereas its ability to generate heat depends on its muscle mass, larger animals are better able to survive a cold winter. This fact led some scientists to suspect that metabolism should be related to mass raised to the power of 2/3. However, empirical work by Max Klieber in the 1930s found a power exponent of 3/4 across a wide range of body sizes.
Geoffrey West went on to explain the common occurrence of the 1/4 factor in many power laws associating physiological characteristics with the size of biological systems. His work suggests that this is because, as they evolved, organisms have been subject to the constraints of living in a 3-dimensional world. The factor, 4, drops out of the analysis, being one more than the number of dimensions.
Two important characteristics are lifetime, which tends to increase in relation to mass raised to the power of 1/4, and heart rate, which is associated with mass raised to the power of -1/4. If you multiply the two together to obtain the total number of heartbeats, the 1/4 and the -1/4 cancel each other out, leaving you with a constant of around 1.5 billion.
According to the NHS, the normal adult heart rate while resting is 60 to 100 bpm, but fitter people have lower heart rates, with athletes having rates of 40 to 60 bpm. Suppose we compare Lazy Larry, whose resting heart rate is 70bpm, with Sporty Steve, who has the same body mass, but has a resting heart rate of 50bpm.
Let’s assume that as Larry eats, drinks coffee and moves around, his average heart rate across the day is 80bpm. Steve carries out the same activities, but he also follows a weekly training plan of that involves periods of elevated heart rates. During exercise Steve’s heart beats at 140bpm for an average of one hour a day, but the rest time it averages 60bpm.
If Larry expects to live until he is 80, he would have 80*60*24*365*80 or 3.36 billion heart beats. This is higher than West’s figure of 1.5 billion, but before the advent of modern hygiene and medicine, it would not be unusual for humans to die by the age of 40.
The key message is that, accounting for exercise, Steve’s average daily heart rate is (140*1+60*23)/24 or 63bpm. The benefits of having a lower heart rate than Larry easily offset the effects of one hour of daily vigorous exercise.
Although it is a rather silly exercise, one could ask how long Steve would live if he expected the same number of heartbeats as Larry. The answer is 80/63 times longer or 101 years. So if mortality were determined only by the capacity of the heart to beat a certain number of times, taking exercise could add 21 years to a lifetime. Before entirely dismissing that figure, note that NHS data show that ischaemic heart disease remains one of the leading causes of death in the UK. Cardiac health is a very important aspect of overall health.
Obviously many other factors affect longevity, for example those taking exercise tend to be more aware of their health and are less likely to suffer from obesity, smoke, consume excessive alcohol or eat ultra-processed foods.
A study of 4,082 Commonwealth Games medallists showed that male athletes gained between 4.5 and 5.3 extra years of life and female athletes 3.9. Although cycling was the only sport that wasn’t associated with longer lives, safety has improved and casualty rates have declined over the years.
Exercise, good nutrition and sufficient sleep are crucial for health and longevity. There’s no point in waiting until you are 60 and taking elixirs and magic potions. The earlier in life you adopt good habits, the longer you are likely to live.
As we move into our 40s, 50s and beyond, we may become aware of changes in our bodies. Performance peaks level off or start to decline. Even if you don’t feel old, it becomes harder to keep up with younger sprinters. It takes longer to recover from a hard ride, injury or illness.
The cause of these age-related changes is a decline in the production of specific hormones. Growth hormone falls insidiously from the time we reach adult height. From the age of 50, testosterone levels drop slightly in men, while oestradiol levels fall dramatically as women reach menopause. The key thing to note about growth hormone and testosterone is that they are anabolic agents, i.e. they build muscle. As they decline, there is a tendency to lose muscle and to increase fat deposition. Sex steroids also play a pivotal role in bone formation.
Fortunately there are measures we can take to counter the effects of declining hormones. Nutrition plays an important role. Understanding the physiological effects of hormonal changes makes it easier to recognise beneficial adaptations in your diet.
Protein provides the building blocks required for muscle. Taking an adequate level of protein, spread out through the day, is beneficial.
Carbohydrates are the key fuel for moderate to high intensity. Fasted training is not advisable. The body’s shock reaction to underfuelled training is to deposit fat.
The UK government advises everyone to take vitamin D supplements, especially over the winter. In addition to supporting bone health, studies have shown improved immunity and muscle recovery.
When combined with adequate nutrition, exercise, particularly strength training, stimulates the production of growth hormone and testosterone. It is important to ensure adequate recovery and to follow a regular routine of going to be early, because these hormones are produced while you are asleep.
Everybody is unique, so you need to work out what works best for you. For further insights on this topic, Dr Nicky Keay has written a book full of top tips, called Hormones Health and Human Potential.
